New Clarinet for MuseScore 2

• Mar 24, 2015 - 14:07

I am nearing the end of entering loop information into the new clarinet I am working on for MuseScore 2 using samples from the University of Iowa.

At this point I need the considered view of the community.

Should the Clarinet only produce notes in the standard range for the instrument?

Or should the range be extended throughout the MIDI keyboard range?

You need to bear in mind that notes outside the official range will have to be timestretched in the lower octaves of the MIDI keyboard, with consequent deterioration in speech and timbre.

The other way forward would be to use Bass Clarinet samples to fill out the bottom of the MIDI range - some of the samples would still need to be timestretched to fill the lowest 3 octaves but sound quality would be better down to C2.

Let me know your thoughts.


Comments

I would say certainly you should go beyond the "standard range," as there are some notes that are almost never played but technically can be produced (on the high end), but notes in what I would call the "fictional range"—notes that actually don't exist—don't really matter.

Bass Clarinet in the low range would be great! I also think it would be a good idea to use an E-flat in the highest range/above the range of the clarinet (I'm pretty sure the UI samples have an E-flat set).
Are you going to be using all of the dynamics samples in the Soundfont or are you only going to be using the mezzo-forte dynamic in your Soundfont?

Don't forget to include samples with sharp attacks. I've had to forego many soundfonts because the attack is too slow, which means that I can't use staccato or sforzando to any real effect, plus trills are next to impossible to set up (the start time and end time hafta be perfect).

Don't know if clarinets ever have a strong attack but it's worth considering...

And thanks for asking our input. I'll keep my open for your finished project! ;)

In reply to by harbinger

The tricky part is finding the right balance between an attack that is too slow (making short notes too quiet) and one that is too hard (making it impossible to create a legato effect). Unfortunately, MuseScore currently provides no way to differentiate the sample used for a slurred or tenuto passage versus one not specifically indicated as legato, nor does the underlying FluidSynth provide a way of making this happen via the legato controller as far as I know (it works only in monophonic mode?). This is something I am hoping can perhaps be addressed in a future release.

In reply to by harbinger

I'm a bit puzzled by your comment about trills.

Trills are not executed with tonguing on woodwind instruments, but by movement of the keys/finger holes on the same stream of air.

So why should attack be relevant????

FWIW I have chosen the mf set of the University of Iowa because it represents a reasonable attack to the note which can be increased or lessened by the soundfont envelope parameters.

The pp set was a) too lacking in attack, and b) not consistent enough in tone quality in some of the registers of the instrument, most notably the chalumeau and top end.

If you'd like to check the samples out yourself, here's the link......
http://theremin.music.uiowa.edu/MISBbclarinet.html

In reply to by ChurchOrganist

Attack is relevant for trills for the same reason as for any other short note - if the attack is too slow, the note does not have a chance to come up to full volume before the "note off" event comes along. This has been a serious problem for a lot of other otherwise excellent soundfonts - staccato notes, 16th notes, grace notes, trills, and other notes with a short play time sound too quiet because the attacks are too slow and the notes never come up to full volume.

Finally I have the loops all done, so I am proud to present the first beta version of the MuseScore Clarinet Soundfont.

Currently it only works for the offical clarinet range so there are no sounds below D3(MIDI#50) and above C6(MIDI#96), and there are no velocity splits yet, and no vibrato

You can get it from my Google Drive........
https://drive.google.com/file/d/0B7cZM0RQwkwSZFVkeHJ5a1hObVE/view?usp=s…

Overall volumes and timbre seem to be good on my sound system.

The main questions I need answering by those of you testing this are...
1. Tuning - are there any notes sounding out of tune?
2. Looping - are any of the loops causing problems? You will find some of them produce amplitude vibrato which is due to amplitude fluctuations within the two loop points. Please let me know of any of these that need tweaking.

The loops deliberately don't start until well into the samples so only very long notes are affected, which would normally have vibrato applied to them anway.

Please also make any other relevant comments.

In reply to by ChurchOrganist

One note that the loops are definitely causing problems with is (in concert pitch) the second A above middle C (the Bb clarinet's B). But I also found myself having the famous "noise gate popping" issue, which I had hoped would not be present in the new clarinet (it isn't present in any other SoundFont I've tried). Unfortunately, whatever it is that causes this in the current FluidR3Mono is also a factor in MSClarinet.

There's also the curious way that some of the notes in the low range seem to have some sort of noise, presumably from the original microphone, like a faint breathing or hissing, while others are perfectly clear.

In general, though, the sound is beautifully clear and realistic, and the tone is a vast improvement. Thank you.

In reply to by Isaac Weiss

I think the hissing you are hearing is actually breath noise, the microphone may have been placed too near the instrument when recording these notes.

It could also be a feature of the clarinet or the particular reed being used.

A friend of mine played clarinet and one feature of the instrument is that not all notes are crystal clear.

Personally I find that such subtleties enhance the impression that it is a real instrument.

I'll check that loop you mention and see what can be done to minimise the problem - it may be that I need to take the loop from a steadier part of the sample.

In reply to by Isaac Weiss

Finally got the chance to check out the sound with headphones. Definitely very nice, and I don't hear "hissing" in low notes - maybe I am not listening to the right notes or for the right things. I also don't hear any "pulsing" issues on long notes any more than I do with any other sounds. But I *do* hear that "noise gate popping" sound as strongly as ever.

Oh - I did notice that low F & F# (written pitch) arer noticeably quieter than the rest of the notes.

One more thing: the samples go down only to the low E, which is fine of course, but it seems we would need to to integrate the lower samples from FluidR3 in order for this sound to work as a replacement - I and I imagine others use the clarinet sound for bass clarinet as well.

In reply to by Marc Sabatella

Adding bass notes to the range will be the next phase of the project - I have yet to pre-process the Bass Clarinet samples so they can be added.

I'm a bit mystified by the gate popping - it seems to be a feature of Fluid's sample handling - Viena/Synthfont doesn't behave like this.

It is excessively annoying so I must work out how to stop it - I'm guessing atm that the key may lie in the release envelope.

Thyere is attenuation on the two lowest notes as they didn't sound balanced - maybe my ears were playing tricks.

In reply to by ChurchOrganist

I'm now thinking that the sound is too quiet. There's two parts to this: one is that the tone sounds like a person playing the clarinet quietly, no matter how amplified (I'm guessing this is not going to change), and the other is that it's simply quieter—noticeably quieter than the flute (easy to change).

Version 0.9 beta is now ready.

https://drive.google.com/file/d/0B7cZM0RQwkwSZFVkeHJ5a1hObVE/view?usp=s…

This (hopefully) fixes the popping gate problem in FluidSynth, and also the inconsistnet volume of the bottom 3 notes.

More news - I have now downloaded the mf Bass Clarinet set from the University of Iowa, and will be starting work on splitting them up into separate samples and normalising them ready for loop finding shortly.

This will give us real samples down to Db2(MIDI #37) - below that they will have to be extrapolated.

There is one duff note so far which we will have to decide whether to keep in for realism or not.

There are also key clicks at the end of some of the samples which might be interesting to turn into a separate effect soundfont.

This clarinet voice sounds sublime at first, but I noticed a few things as I listened on. The reverb is very noticeable, which isn't a true hindrance to the ultimate quality of the sound; especially fantastic in some largo settings, but can be somewhat distracting, especially when used with faster, more aggressive pieces, such as Bizet. This font is ideal for typical concert pieces, however.
The only other problem I have with this voice is the fact that the low notes are very weak. This isn't too bad for say, first part clarinetists, but as a bass clarinetist myself, it's tearing me apart to not hear the crisp tones of the lower notes.
To summarize, this font is dreadfully good for use in high octave soprano clarinet parts in slower movements, but work needs to be done on the lower parts. Very nice.

In reply to by ATMSPHR

Thank you for your comments.

There is no reverb on the samples - they were recorded in an anechoic chamber, so any reverb would be added by MuseScore's Zita reverb.

I am actually currently working on adding Bass Clarinet samples to the soundfont which will extend the available range down another 2 octaves, and also provide a Bass Clarinet to the soundfont.

I shall also be eventually adding samples with a more aggressive attack.

In reply to by ChurchOrganist

The sound is awesome! Adding the bass clarinet will be a big help. I use the clarinet samples to replace the choir sounds sometimes (which are awful) and once the bass clarinet is added it will be perfect for this because it has such a pure tone.

There aren't any samples on the website, but there is such a thing as a contrabass clarinet. Maybe someone can provide samples for that, too. If anyone has one, I bet they're really expensive. And of course it would be difficult to match University of Iowa's sound, but it would be nice to add that EVENTUALLY.
Also, Is it a long-term goal to use all/many of these sound samples to replace the current MS sounds?

I'll try to use this soundfont and listen for anything that needs fixing, in the meantime, great work and thank you!

In reply to by ChurchOrganist

But is it really worth doing? Haven't we come to the conclusion that almost all SoundFonts are inferior to FluidR3, and shouldn't we therefore assume that odds are that our best efforts will probably still not sound as good?

I do have issues with certain instruments in FluidR3 messing up the ensemble sound—alto saxophone being a prime example, a certain violin range being another, and the clarinet being a third—but overall it is more realistic than even GeneralUser—less in tune in certain instruments, but more realistic.

I'm strongly in favor of improving FluidR3, but I think starting from scratch is unlikely to be successful.

In reply to by Isaac Weiss

I think there is *considerable* disagreement about soundfonts. Not everyone likes FluidR3. There is value in having a soundfont where all instruments were recorded with the same equipment, under the same conditions, etc. In theory, you might get a more harmonious blend. But it's a huge undertaking, to be sure.

In reply to by [DELETED] 5

The SFZ format seems to be easier to work with (plain text, audio file) than SF2.

Actually no.

The big problem with SFZ is that there is no decent graphical editor.

There is sfZed but that is nearly as bad as writing text files, and has had no development since 2010

It is far easier to produce a soundfont with Polyphone or Viena as they contain tools to make region and split editing, plus there is a degree of onboard sample editing, looping audition tools etc all of which are currently missing in an SFZ environment.

There is a new graphical SFZ editor but I haven't put it through its paces yet.......
http://patcharena.com/tag/free-sfz-editor/

In reply to by Isaac Weiss

Yet starting from scratch is exactly what I have done with the clarinet.

Provided you understand sound design and have an editor such as Viena or Polyphone and access to a decent set of samples the process is relatively simple.

Time-consuming - but still simple.

PS and there are many better soundfonts than FluidR3

In reply to by ChurchOrganist

About making a MS soundfont (more-or-less from scratch, based on FluidR3):

I think it would be great if we could pull together U of Iowa's professional sounds and maybe some sounds provided by Musescore's community (many of whom are talented musicians) and make a professional quality soundfont that ships with Musescore. Or even, include FluidR3 these individual sounds as an extension/upgrade to FluidR3. But: it doesn't help the notation goal any. It's great for playback (and since I use Musescore primarily to write music, I would love that) but doesn't add anything to Musescore's notation abilities (which I understand is the main goal. Is there a mission statement or something explicitly outlining the goals that someone could link me to?)

I think that making soundfonts for each of these samples is a good idea that should probably be community led and leave the developers to focus on other tasks.

I think that I can supply solo-voice (male) samples, ukulele samples, steel-string guitar and electric guitar samples (clean, distorted, anything, I use a digital amp), but I can't guarantee that the electric guitar samples will be consistently in tune. If someone sends me links to tutorials and stuff, I might be able to help make soundfont presets (school just ended so I have a lot of time on my hands).

Anyways, if we are going to seriously consider or pursue a long-term soundfont upgrade like this it should probably have it's own forum topic! In the meantime, thanks for the awesome work on the clarinet and good luck adding bass clarinet samples!

PS and yeah, I use Timbres of Heaven, which has more balanced volume IMHO, and the voice oohs are good enough to replace the SUCKY aah choir. One of these days I am going to get samples from a good choir and make a soundfont to replace the aaahhful choir.

In reply to by joseph.branden…

Go to the first (actually last) page of this forum and you'll find a number threads discussing creating our own soundfont. It was indeed one of the main reasons this forum was created. But it was quickly realized the project bigger than anyone had time for.

I'm still wondering about the possibility of licensing Timbre of Heaven, in whole or in part. It's the only soundfont I know of that I can would consider an improvement overall. Of course, I only know the free ones.

In reply to by Marc Sabatella

Actually IIRC the main problem was the lack of donated samples.

Of course soundfont creation is time-consuming, but then so is programming, and once you have got your head round the principles involved the process is actually quite repetitive.

As the key to a successful soundfont is the quality of the samples used the lack of suitable donations was the main cause of nothing happening, coupled with focus on actually getting MuseScore 2 finished.

Furthermore playback improvement wasn't particularly on the agenda back in 2012 as the problems in FluidR3 were yet to be recognised.

In reply to by ChurchOrganist

Re: Starting from scratch

What I meant was that it seems like a bad idea to try to build a full GM SoundFont from the beginning. I would definitely support more work similar to the clarinet project—i.e, starting with FluidR3 as a base, and replacing one instrument at a time (ideally in order of necessity, with the overwhelming majority of the sounds probably remaining).

The idea is that a) if the project is never finished (or for that matter at any given stage prior to the project being finished) the work won't have come to nothing, and we'll be able to take advantage of it; and b) it cuts down on the workload considerably if we can just agree that, say, the acoustic guitar sound in FluidR3 is great as it is.

The Bass Clarinet samples are in but no loops yet.

There are now 2 presets in the soundfont both assigned the GM Patch number for Clarinet #72.

Clarinet is in Bank #0 and Bass Clarinet in Bank #1

Perhaps you would be good enough to listen carefully to the tuning of Bass Clarinet as Viena's auto tuning mechanism freaked out on a few of the samples and gave spurious results for note number, so it is probable that tuning is not spot on for these samples and will need tweaking manually.

As ever the link is.....
https://drive.google.com/file/d/0B7cZM0RQwkwSZFVkeHJ5a1hObVE/view?usp=s…

In reply to by Isaac Weiss

Could this be a Mac OS X thing? Are there any other Mac users experiencing this??

It's working fine in MuseScore on Windows 10 - it is a little quiet, but by no means inaudible.

Do you have another soundfont player on your Mac which you can test with?

Does Garageband support soundfonts?

In reply to by ChurchOrganist

I loaded the SoundFont into Polyphone, and the individual samples do play (and I found a couple that don't seem to be very well looped, but that's beside the point). Is that what you meant by "test player"?

Also, GarageBand only works with its own samples.

Anybody else with a Mac to see if this works?

In reply to by Isaac Weiss

Did you try playing the presets with Polyphone's test player?

Polyphone has a virtual keyboard used for test playing soundfonts.

There are one or two dodgy loops - unfortunately the nature of loop finding is imperfect - you cannot always get a perfect match in the waveform, but I will be tightening this up in due course using Viena's loop smoothing tool.

Yes, I thought that was the case with Garageband - definitely entry level stuff, and not meant for professional use.

So we just need to track down why it's not working in MuseScore

In reply to by Isaac Weiss

So why silence in MuseScore I wonder?

As I said earlier - when first loaded a soundfont doesn't always "take" straight away.

I find I sometimes have to do a few patch changes in the mixer before I get the soundfont to play.

Maybe that is what you are experiencing?

Always difficult to solve these problems at long range :)

In reply to by ChurchOrganist

Well, I messed around with adding and removing SoundFonts and changing patches in the Mixer, and after a few tries it started working. Upon subsequent attempts, it seems to reliably make no sound until the patch has been changed at least once, which I don't recall ever experiencing before. I like the sound, though—it's been worth the time invested.

In reply to by ChurchOrganist

I've just installed the new clarinet sound (shows as "Clarinet1" in the Mixer), and it's a big improvement over the previous sound, which always seemed overly reedy and pinched. If anything, the new sound goes in the other direction: more soft and flute-like. Also, the volume is lower, and I had to push it to the max in the Mixer, about 4:30 o'clock on the dial where the flute, oboe and bassoon are at 1:30.

When I installed the new clarinet soundfont in the Synthesizer, all the instruments in the score I had open reverted to Clarinet1 in the Mixer, and I had to go through and reset them individually. Is this the expected behavior?

I saw the note about doing "Save to Score" and "Load from Score" in the Synthesizer before closing and after reopening the score respectively. Is this also normally required?

In reply to by dhfx

Addition to my previous post: When I went to "Save to Score" in the Synthesizer, I found the buttons grayed out. After I saved and reloaded the score and clicked "Load from Score" I found the instrument selections in the Mixer were changed and I had to reset them individually again. Then when I went back to the Synthesizer I found "Save to Score" grayed out again.

BTW I'm using MS 2.0.0 on Linux Mint 17.1 - is this a bug that's been fixed?

In reply to by dhfx

I don't think what you are describing is a bug. As far as I know, the clarinet soundfont posted was for testing purpsoes only and not intended to for production use yet. But I suspect if you add it to the *end* of the list of loaded soundfonts, then other instruments will retain their original sounds, and then you'd just have to select the new clarinet sound where desired in the Mixer.

And I also think it is normal that save / load buttons would be greyed out if there is nothing to save or load. That is, after saving or laoding, until you make a change to the soundfont list, there is nothing else to save or load. The options become available once you modify the soundfont list,

In reply to by Marc Sabatella

Indeed it is still in Beta.

The Bass clarinet samples still need to have loop points ascertained, and I have yet to look at the overall volume balance of the soundfont in terms of db level output.

All I can say is watch this space - once I have the scoring finished for a musical production in October, then I will have a little more time to work on it, although I'm trying to work on an improved Sounds menu for the mixer too.

In reply to by ChurchOrganist

The E flat clarinet is now in, albeit without looping. Range is limited to the official E flat clarinet sounding range ie G3 to C7.

So I'm pleased to present MSClarinet 2.00 beta

I'm currently not convinced that there is enough of a difference between this and the B flat version to justify its inclusion given that the uppermost limit of the range is the same as the B flat.

Please let me have your comments.

Just in case you need the link again - here it is......
https://drive.google.com/open?id=0B7cZM0RQwkwSZFVkeHJ5a1hObVE

In reply to by ChurchOrganist

whoa, tried your MSClarinet latest soundfount and it sounds very, very good!

but the bass clarinet..the looping is finalized? I'm attaching (again :-)) a test score for you to check, and a capture of the audio of the Musescore playback (playback button) of it. The Clarinet sustains the sound for all the dotted quarter/half notes , but the bass clarinet seems to lag behind

I mentioned that a number of the samples seem poorly looped—how should I go about reporting these?

Also, studying the file through the lens of Polyphone, I'm surprised you seem to have separate samples for every semitone. Might that not be necessary?

In reply to by ChurchOrganist

Let's just start with the Bb Clarinet. Some of these are worse than others (e.g., 81 is much more glaring than 80, and 82 and 83 are fairly subtle), but all of the following have looping problems: 50, 53, 59, 60, 63, 65, 68, 72, 73, 74, 76, 80, 81, 82, 83, 84, 86, 89, 90, 91, 92, 93, 94.

I asked about the number of samples because it occurred to me that it could cut down drastically on the creator's workload with no distinguishable affect on output quality—from what I've seen, the most realistic SoundFonts have a sample for every two or three semitones, and it's absolutely impossible to tell the difference from the top of one sample's range to the bottom of the next.

There's also the fact that not all of the samples are actually good, so to maintain consistent quality you might be forced to skip some of them and let their neighbors fill in. For example, 50 is very muddy while 51 is a clear tone; 55 comes with a hiss of air that's not present in 54 or 56; 66 is undeniably ugly; and 71 is decent, but sounds like a different instrument than the samples on either side.

I sincerely hope I'm not coming off as hyper-critical. It's just that I play clarinet, so I want a good clarinet sound. ;-)

In reply to by Isaac Weiss

Thing is Zack, you almost never get a perfect loop.

This is because the nature of the waveform is so complex that when you repeat part of it it doesn't fit quite right.

In most of these instances, the loop comes over as a vibrato, and is unlikely to be noticed when not played in isolation. There are one or two which do need looking at, but I will get to those when I have finished defining loops for the E flat clarinet.

The imperfections in the samples are probably a feature of the instrument.

No clarinet plays exactly the same as another, and also no clarinettist will blow the same note exactly the same twice. It is these imperfections which add to the realism of the soundfont, and again when not heard in isolation they are not so noticeable.

The clarinet has 3 defined registers with a definite change in timbre between them, so a clarinet soundfont which has a consistent sound throughout its range is not actually a very good one.

Thank you for creating this soundfont! I've been searching for ages for a decent clarinet soundfont so I could make a videoscore for my transcription of 'The Tale of Viktor Navorski,' which is the main theme from the move 'The Terminal,' scored by John Williams. The piece is a clarinet solo, so it was pretty important to find a good soundfont for it. Thank you again, and please listen to my videoscore, which I think showcases your handy-work quite nicely, if I may say so.
https://musescore.com/node/1712451

In reply to by trumpetdude315

It certainly does, and the contrast between that and the default clarinet reveals what a joke the default is :)

BTW the player on the link in your post isn't working. I had to go to Youtube to listen to it.

Thank you for doing this. I shall put it on the list of demos for the soundfont when it is finally out of beta.

After some careful study, I decided to revise the score. I made it a few years ago, so the formatting and notation wasn't quite up to my current standard. I also had to redo the videoscore, so here's the link to the sheet music page, and you can go to the videoscore from there. That way, if I ever change the videoscore again, I won't have to change this link.
https://musescore.com/user/23822/scores/134368

In reply to by ChurchOrganist

The only criticism I've got for this is that it doesn't have enough dynamic range - piano and fortissimo don't sound much different. That's fine for oboes, but clarinets are drama queens that can go from whisper quiet to nearly french horn loud. I'd suggest adding a velocity layer, but I'm sure there has to be another solution. Sound character and attack don't change a heck of a lot for clarinets, so you shouldn't need more than one layer.

In reply to by ChurchOrganist

A few observations: The volume of the new clarinet samples is considerably lower than the old FluidR3 clarinet. Also the F# above middle C sounds too breathy. And in the bass clarinet the attack seems too slow on the A below middle C.

(I'm referring to the notes as played on the on-screen keyboard).

Hello! So, I discovered something that might make your life easier... or make you want to punch a wall, I don't know.

The Sonatina Symphony Orchestra Clarinet (SSO Clarinet) *is* the UoI Clarinet. So If you wanted to use that, you wouldn't need to do any work at all.

Also, about the volume level - I find it better to normalize all the .wav files before I import into an sf2.

Do you still have an unanswered question? Please log in first to post your question.